ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Changes
Revision: 1.497
Committed: Mon Jun 22 11:25:53 2009 UTC (14 years, 11 months ago) by root
Branch: MAIN
Changes since 1.496: +10 -4 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 Revision history for Perl extension Coro.
2
3 TODO: should explore PerlIO::coroaio (perl leaks like hell).
4 TODO: maybe implement a default message channel, very much like Erlang's
5 actor model (which is cool in a lot of important aspects (failures!),
6 but very lacking in others (higher level ipc)).
7 TODO: unready_all
8 TODO: myhttpd header parsing
9 TODO: channel->maxsize(newsize)?
10
11 - provide explicit functions to "cede" to the event loop in Coro::AnyEvent,
12 as this seems to have been a difficult concept (poll, sleep, idle,
13 idle_upto).
14 - add Coro::AnyEvent::readable/writable functions.
15 - clarify Coro::EV/Event/AnyEvent manpages.
16 - free per-thread global scalars in the thread calling ->cancel, to
17 avoid crashes when $_, $@ etc., are magical but some of those
18 had already been freed.
19 - speed up ready queue management/context switching by using a linked
20 list instead of an array (~5-10%).
21 - implement "watch" command in Coro::Debug shells.
22 - for fun, implement time-slicing as an example in the manpage.
23 - if AnyEvent detects EV or Event, but we don't have Coro::EV or
24 Coro::Event, use the normal AnyEvent handling instead of dieing
25 (the same is true for Coro::Handle).
26 - properly document Coro::EV::timed_io_once.
27 - avoid unneccessary ->cancel alls in Coro::Handle.
28
29 5.132 Fri May 29 09:00:39 CEST 2009
30 - do not keep a reference to the argument itself in
31 Coro::Semaphore::guard, as it could change later.
32 - support SO_RCVBUF/SO_SNDBUF nonstandard Coro::Socket options,
33 should support a prepare callback.
34
35 5.131 Mon Mar 16 23:20:37 CET 2009
36 - implement and document Coro->suspend, Coro->resume.
37 - fix Coro::Select implementation to not (often) close
38 the passed file descriptors (testcase provided by pippijn).
39
40 5.13 Mon Dec 15 21:51:42 CET 2008
41 - EXPERIMENTAL: implement dynamic winds (on_enter/on_leave).
42 - don't set diehook to C<undef> but instead to NULL, to avoid
43 spurious warnings.
44 - fix a lot of bugs in Coro::SemaphoreSet.
45 - Coro::SemaphoreSet will less often create a semaphore needlessly.
46 - add Coro::SemaphoreSet::count and wait methods.
47 - take advantage of the new Guard module.
48 - deprecate Coro::guard.
49 - try to fix the dreaded 01_unblock tests once more. I hate it when
50 testsuites need more fixing than the code they are supposed to test.
51 - croak in more cases when a required callback isn't resolvable.
52 - fix some minor issues in Coro::State->call/eval.
53 - use current coroutine context instead of a temporary one
54 when temporarily switching to another coroutine.
55 - do not call C-level on_destroy handlers during global destruction,
56 to avoid needless segfaults.
57
58 5.12 Sun Dec 7 13:30:38 CET 2008
59 - add default config for MirOS, which seems to be a bug-to-bug
60 compatible fork of openbsd ("world domination by releasing
61 openbsd cvs before the openbsd folks do it" or so :).
62 - free_padlist did destroy the names pad, not good, but didn't
63 seem to bother perl - this could fix issues such as eval ""
64 inside a function called from multiple coroutines.
65 - use a different method to detect destruction time.
66 - be more careful when freeing padlists just before global
67 destruction.
68 - fixed and expanded the call/cc example.
69 - renamed _terminate to _coro_run.
70 - new method Coro::Channel->shutdown.
71 - try pthreads on openbsd <4.4 (broken sigaltstack, will
72 pthreads fare better?).
73 - be less picky about destroying "a" running coroutine.
74
75 5.11 Tue Nov 25 21:49:05 CET 2008
76 - DEBUGGING in 5.10.0 is a mess: it constantly flags perfectly
77 working code with failed assertions, introducing more bugs than
78 it fixes, requiring elaborate workarounds :(
79
80 5.1 Mon Nov 24 08:54:59 CET 2008
81 - wrote a small introductory tutorial - Coro::Intro.
82 - convert Coro::Timer, Coro::Select and Coro::Util to rouse API.
83 - Coro::Select did errornously dup the file descriptors
84 and didn't work with all AnyEvent backends.
85 - Coro::Select wasn't imported correctly form Coro::LWP, causing blocking
86 LWP data transfers.
87 - disassociate c contexts from coro objects - this is agruably more
88 correct, but mostly allows sharing of cctxs between coro and state
89 objects, for added memory savings and speed increases.
90 - bumped $Coro::POOL_RSS up to 32kb by default.
91 - no longer set the optype to OP_CUSTOM, as B::* understandably
92 doesn't like this very much (and we *are* a type of entersub).
93 - implement state cloning, just to prove that call/cc can be done.
94 - automatically load Coro::AnyEvent in Coro::Handle.
95 - wrap ->cancel calls in eval inside Coro::Handle as EV watchers
96 do not have this method (and don't need it either).
97 - speed up generic anyevent methods in Coro::Handle by using rouse
98 callbacks.
99 - allow coroutines in $Coro::IDLE, speeding up Coro::AnyEvent and
100 others. It also makes the debugger happier, as you can trace
101 through the idle threads now.
102 - add comppad_name* and hints ($^H, %^H) to per-thread variables.
103 - eg/event was pretty broken.
104 - better 5.8.6 compatibility.
105
106 5.0 Thu Nov 20 10:35:05 CET 2008
107 - NEW ARCHITECTURE: use the latest 4.x version if you experience
108 stability issues.
109 - bump API version to 7 - all dependents must be recompiled.
110 - removed timed_* functions - they were not being used anyways
111 and should be replaced by a more generic mechanism -
112 and were annoying to support anyways :)
113 - removed SemaphoreSet's waiter method - use sem method instead.
114 - Coro::Semaphore->adjust didn't correctly wake up enough waiters.
115 - async_pool did free a scalar value twice
116 ("Attempt to unreference...").
117 - fix a longstanding bug where calling terminate on a coro that
118 was waiting for a semaphore that was just becoming available
119 would cause a deadlock (semaphore would get into a state where
120 it was available but waiters were still blocked).
121 - calling throw on a coroutine that is waiting for a semaphore will
122 no longer make it acquire the semaphore (and thus leak a count).
123 - perl's process emulation is now not even theoretically supported
124 anymore.
125 - new functions Coro::rouse_cb and Coro::rouse_wait for easier
126 conversion of callback-style to blocking-style.
127 - new methods $coro->schedule_to and ->cede_to, to specifically
128 schedule or cede to a specific coroutine.
129 - new function Coro::Semaphore::wait.
130 - use named constants in Coro::Channel (Richard Hundt).
131 - directly patch the entersub opcode calling SLF functions (cede,
132 transfer and so on). this does speed up context switching, but
133 more importanly, it frees us from the hardcoded behaviour of
134 entersub, so we might actually be able to return something from
135 those functions and atcually create new ones.
136 - take advantage of __builtin_frame_address on gcc.
137 - expose THX in coroapi (not sure whether this was a wise decision,
138 as "threaded" perls are running at half speed anyways).
139 - implement execute_slf (schedule-like-function) interface that makes
140 it possible to implement schedule-like-functions in XS.
141 - use new SLF interface to massively speed up Coro::EV by roughly a
142 factor of two.
143 - used new SLF interface to massively speed up Coro::Semaphore by a
144 factor of three.
145 - used new SLF interface to speed up Coro::AIO by roughly a factor of
146 four and reduce its memory usage considerably.
147 - implement Coro::SemaphoreSet purely in terms of Coro::Semaphore,
148 for a nice speedup and vastly more correct behaviour. Also implement
149 a new method "sem" to get at the underlying semaphore object.
150 - implement Coro::Channel in terms of Coro::Semaphore, for a moderate
151 (in comparison) 20-40% speedup.
152 - used new SLF interface to reimplement Coro::Signal gaining
153 some unknown (because I was too lazy), but certain, speedup, and also
154 making signals reliable for the first time.
155 - used new SLF interface and other optimisations to speed up async_pool
156 by a factor of two. It also doesn't rely on perl's exception mechanism
157 to exit anymore. The overhead for terminating an async_pool, coro over
158 a normal async is now very small.
159 - sped up coroutine creation/destruction by 40%.
160 - forgot to include Coro/libcoro/README in the dist for all these years.
161 - work around a freebsd pthreads bug (manual testcancel is required as
162 pthread_cond_wait isn't a cancellation point on freebsd).
163 - use new rouse functions to speed up and simplify Coro::BDB.
164 - make "prefer perl native functions" work with threaded perls.
165 - condense Coro::Debug ps output, hint at v and w flags.
166 - (libcoro) lots of minor cleanups and portability improvements.
167
168 4.914 Wed Nov 19 12:54:18 CET 2008
169 - fix a disastrous bug in the readline optimisation
170 introduced in 4.801.
171
172 4.913 Sat Nov 15 07:58:28 CET 2008
173 - async_pool did free a scalar value twice
174 ("Attempt to unreference...").
175
176 4.912 Thu Nov 13 18:31:23 CET 2008
177 - minor cleanups.
178 - use much larger stacks on linux and perl < 5.8.8.
179 - Coro::Debug::new_unix_server did not unlink the socket
180 when destroyed.
181
182 4.911 Tue Nov 11 04:26:17 CET 2008
183 - "port" to threaded perls.
184
185 4.91 Mon Nov 10 05:36:38 CET 2008
186 - the ->throw exception object no longer leaks.
187 - creating a new cctx leaked a scopestack entry (memleak).
188 - new coroutines didn't get created with a zero flags field
189 (unknown impact).
190 - calling ->throw on a not-yet-started coroutine should now work
191 instead of being ignored.
192 - ->throw is now supported on Coro::State objects.
193 - clean up cctx creation code a bit.
194 - entersub is actually a UNOP, not a LOGOP (not a bugfix).
195
196 4.9 Sat Nov 8 17:45:27 CET 2008
197 - (libcoro) did not preserve rbp with CORO_ASM (we are getting there).
198 - (libcoro) no longer leak threads in the experimental pthread backend,
199 also speed it up considerably.
200 - (libcoro) do not rely on makecontext passing void *'s unscathed.
201 - fix compiletime dependencies on libcoro in the Makefile.
202 - cctx_count wasn't always updated properly.
203 - Coro::State::cctx_stacksize wasn't applied correctly.
204 - new function Coro::State::cctx_max_idle.
205 - the default max number of idle C contexts is now 4.
206 - (libcoro) try harder to get _setjmp/_longjmp.
207 - (libcoro) cleanup and extend the libcoro API to officially
208 allow the creation of empty source contexts.
209 - very experimental workaround for the totally broken netbsd platform.
210 - tried to hack around openbsd bugs.
211
212 4.804 Wed Nov 5 16:36:00 CET 2008
213 - Coro::Debug::new_unix_server would not create a non-blocking listening
214 socket, sometimes causing freezes.
215 - (libcoro) fix misaligned stack points for setjmp and assembly
216 methods, which can cause crashes on x86/x86_64 with a sufficiently
217 aggressive compiler.
218 - new function: Coro::Debug::new_tcp_server.
219 - move ->throw into the Coro class because it only works on coro objects.
220
221 4.803 Mon Nov 3 17:16:12 CET 2008
222 - (libcoro) use a global asm statement to become independent of gcc
223 otpimisations for CORO_ASM (thanks to pippijn for the idea).
224 - try to workaround yet another broken bsd, this time dragonfly.
225
226 4.802 Thu Oct 30 10:56:12 CET 2008
227 - support -fno-omit-frame-pointer on x86 with the assembly method.
228 - tune 01_unblock.t tests a bit.
229
230 4.801 Wed Oct 22 18:33:37 CEST 2008
231 - improve readline speed for very long "lines".
232 - backport to 5.8.8.
233
234 4.8 Thu Oct 2 13:34:40 CEST 2008
235 - new function Coro::AIO::aio_wait.
236 - Coro.:AIO and Coro::BDB now "use Coro::AnyEvent".
237 - greatly speed up and reduce memory usage of Coro::AIO requests.
238 - implement some other ยต-optimisations.
239
240 4.749 Mon Sep 29 14:40:12 CEST 2008
241 - port to slow and broken pseudo-threaded perls. (courtesy pippijn).
242
243 4.748 Sat Sep 27 14:03:19 CEST 2008
244 - implement, but do not document, PerlIO::cede(granularity).
245 - Coro::Storable forgot to wrap Storable::pstore.
246 - work around the multitude of leaks and memory corruption
247 bugs in PerlIO::via by using our own C-level perliol. As a side
248 effect, Coro::Storable is now much, much, much faster.
249
250 4.747 Tue Sep 23 01:59:41 CEST 2008
251 - fix a per-cv memleak (one empty array was leaked per
252 code reference).
253 - avoid a crash in coro->call|rss when the coroutine was already
254 destroyed (most noticably when using Coro::Debug::ps :)
255 - also protect *Storable::FILE.
256 - push up default storable granularity to 20ms.
257
258 4.746 Sun Sep 21 03:22:20 CEST 2008
259 - be more insistent on locking Storable against reentrancy
260 in Coro::Storable.
261 - move swap_def?v and throw to Coro::State, as documented.
262
263 4.745 Thu Jul 24 00:14:38 CEST 2008
264 - remove debugging code related to MgPV_nolen_const, also try to
265 make it compile with perl 5.8.6 (yes, apple apparently loves
266 outdated software). Reported by John S.
267
268 4.744 Tue Jul 8 22:06:35 CEST 2008
269 - correctly provide default DIE/WARN handlers as documented.
270 - also overwrite PL_vtbl_sigelem.svt_clear, even though current
271 implementations inside perl work fine for us.
272
273 4.743 Mon Jun 16 00:21:57 CEST 2008
274 - when using Coro::EV without running EV::loop it could
275 result in busy-waiting for events, this has been fixed.
276 - reduce codesize and improve performance by using EV_DEFAULT_UC.
277
278 4.742 Sat May 31 14:10:21 CEST 2008
279 - implement a workaround for (some) perl <5.8.8 versions.
280 - require EV 3.3+.
281
282 4.741 Fri May 30 23:33:09 CEST 2008
283 - tell netbsd how utterly broken their imitation of an OS is
284 and refuse to build by default if pthreads are in use.
285 - switch to "s" method on all bsd's by default, as their ucontext
286 stuff seems just too broken.
287 - fix a bug in Coro::Select.
288
289 4.74 Thu May 29 20:05:31 CEST 2008
290 - do not test Coro::LWP for lack of dependencies.
291 - do not test Coro::Select for lack of working perls.
292
293 4.73 Thu May 29 2008
294 - fix a bug in Coro::EV which would cause it to block despite
295 there being runnable coroutines.
296 - sprinkle "no warnings" freely over everything, also suppress
297 warnings for some other modules.
298 - fix typo in WSAEWOULDBLOCK.
299
300 4.72 Sun May 25 05:14:36 CEST 2008
301 - tweak META.yaml a bit, unfortunately, there is no documented way
302 to have optional dependencies with CPAN. doh :(
303 - avoid running some tests on windows because they would fail due to
304 perl bug (broken fork, broken pipes...).
305 - work around perl on windows bugs where perl returns undocumented
306 error codes for sysread, syswrite etc. by taking advantage
307 of AnyEvent's workaround for that problem.
308 - use AnyEvent::Util::fh_nonblocking in Coro::Handle to work around
309 a common perl implementation bug on windows.
310 - use unix domain sockets in testsuite to work around
311 common perl implementation bugs on widows (they are emulated by
312 tcp sockets on windows. ugh.)
313
314 4.71 Sat May 24 20:01:27 CEST 2008
315 - fix a bug in Coro::AnyEvent ("Usage: Coro::AnyEvent::_schedule()").
316 - take advantage of async name resolution of AnyEvent::Util.
317 - work around brutal inet_aton override in Coro::LWP.
318 - take advantage of the readyhook in Coro::EV, for smoother
319 scheduling.
320
321 4.7 Sun May 11 00:32:19 CEST 2008
322 - completely reworked the Coro manpage.
323 - added Coro::AnyEvent, generic event loop integration.
324 - implement cancel, ready and kill commands in Coro::Debug.
325 - document find_coro in Coro::Debug.
326 - incompatible change: rename has_stack to has_cctx.
327 - Coro::AIO and Coro::BDB no longer force event model detection,
328 use AnyEvent::AIO and AnyEvent::BDB.
329
330 4.6 Sat Apr 26 10:05:14 CEST 2008
331 - INCOMPATIBLE CHANGE: sub/code attributes are no longer supported
332 by the Coro module. It was a mistake to have it in the first place.
333 - (experimental) support for activestate perl 5.10 (method "w").
334 - (experimental) support for strawberry perl (method "a").
335 - coro_sigelem_set did not return a value although it had to,
336 actual impact unknown.
337
338 4.51 Mon Apr 14 13:28:27 CEST 2008
339 - make it compile again on 5.8.
340
341 4.50 Thu Apr 10 09:43:17 CEST 2008
342 - I did it twice! (see 4.49).
343
344 4.49 Sun Apr 6 21:23:31 CEST 2008
345 - grr, instead of compiling the recent changes on 5.10 only they
346 were compiled on 5.8 only.
347
348 4.48 Sun Apr 6 20:36:46 CEST 2008
349 - allow coroutine switches during eval's under 5.10.x, as it seems
350 the parser is a per-interpreter option now, so this is safe
351 (this might fix the odd crashing bug).
352 - drop support for 5.9.x versions: they are dead, jim.
353
354 4.47 Sun Apr 6 00:37:52 CEST 2008
355 - force cctx allocation on API calls: we know we need to force one
356 and gcc actually manages to confuse our heuristic nowadays,
357 leading to crashes and worse.
358 - document force_cctx.
359
360 4.46 Fri Apr 4 22:05:43 CEST 2008
361 - upgrade libcoro, resulting in pthread-backend (which was only created
362 to fulfill the rules of the programming languages shootout).
363
364 4.45 Thu Mar 13 11:55:36 CET 2008
365 - fix a file leaking bug in eg/mhyttpd that would allow
366 downloading of any file (reported by oesi).
367 - fix deadlock bug in Coro::Channel (reported by Richard Hundt)
368 (also add testcase).
369 - support Broadcast option in Coro::Socket (patch by Richard Hundt,
370 apparently having loads of fun with that).
371 - implement $state->swap_defsv, ->swap_defav and document ->throw.
372
373 4.4 Wed Feb 13 16:44:29 CET 2008
374 - only assume .cfi pseudo isns exist on GNU/Linux.
375 - add get_timed method to Coro::Channels.
376 - fixed Coro::Debug manpage.
377 - perl 5.11 compatbility improvement based on patch by
378 Andreas Kรถnig.
379
380 4.37 Sun Jan 20 11:25:23 CET 2008
381 - soften the check for an active parser for perl 5.10 (avoiding
382 "Coro::State::transfer called while parsing" in many cases).
383
384 4.36 Sun Jan 13 10:53:56 CET 2008
385 - reset diehook when terminating from an async_pool as to not
386 trigger any __DIE__ handlers.
387
388 4.35 Sun Jan 13 04:14:13 CET 2008
389 - "bt" debug command now displays any exceptions
390 from longmess and also skips the innermost
391 stackframes, giving more useufl output.
392 - allow backtraces in coroutines blocked in Coro::EV,
393 at a <1% speed hit.
394 - handle localising of $SIG{__WARN__} and __DIE__
395 properly (with a proper amount of dirty hacking).
396
397 4.34 Sat Dec 22 17:49:53 CET 2007
398 - upgrade to EV version 2.0 API.
399
400 4.33 Mon Dec 17 08:36:12 CET 2007
401 - make Coro::AIO etc. loadable in the absence of EV.
402
403 4.32 Mon Dec 17 07:46:02 CET 2007
404 - majorly improved Coro::Handle's performance with Coro::EV.
405 - implemented the until now mythical Coro::BDB module.
406 - specialcase EV in Coro::AIO and Coro::BDB for extra speed.
407
408 4.31 Wed Dec 5 12:32:39 CET 2007
409 - remove warn statement form Coro::Util.
410
411 4.3 Tue Dec 4 20:33:14 CET 2007
412 - calls to the idle function could cause stack corruption
413 when the stack changed.
414 - do no longer rely on the presence of EV::DNS if EV is used
415 (because it is gone), but instead take avdantage of EV::ADNS
416 if available.
417 - add ($) prototypes to all functions in Coro::Storable.
418 - use a conventional (safer) idle callback in Coro::EV.
419 - do accept quickly in Coro::Debug to avoid endless loops.
420
421 4.22 Fri Nov 30 16:04:04 CET 2007
422 - really use optimised versions for Event and EV in Coro::Util
423 and Coro::Handle.
424
425 4.21 Sun Nov 25 10:48:59 CET 2007
426 - fix a spurious memory read.
427 - Coro::EV no longer keeps the eventloop "alive".
428
429 4.2 Fri Nov 9 20:47:05 CET 2007
430 - enable/disable tracing from a new coroutine, not a pooled one.
431 - fix a memleak in Coro::Event.
432 - removed killall call from fork_eval.
433 - made sure store_fd is already loaded so that fork_eval
434 does not have to parse autoload in each subprocess.
435 - only use assembly method if -O switch is in $Config{optimize}.
436 - add (optional) Coro::EV module, so far the best event loop module
437 directly supported by Coro.
438 - if the event model is EV, use EV::DNS to resolve
439 stuff in Coro::Util.
440 - don't get confused by multiple event notifications in Coro::Handle.
441 - initial support for EV (libevent interface).
442 - require Event and EV using configure_requires, to force their existance.
443
444 4.13 Wed Oct 24 07:26:45 CEST 2007
445 - add Coro::Storable::blocking_thaw.
446 - use a vastly more complicated technique to localise
447 $SIG{__WARN/DIE__} that also works on perls <= 5.8.8.
448 - use a coroutine for the idle callback Coro::Event,
449 instead of running Event in the current coroutine context.
450 This also catches recursive invocations.
451 - actually report fork errors in gethostbyname and inet_aton.
452
453 4.11 Thu Oct 11 02:40:24 CEST 2007
454 - port to threaded perls.
455
456 4.1 Thu Oct 11 02:38:16 CEST 2007
457 - incompatible change: $SIG{__DIE__} and $SIG{__WARN__} will now
458 be local to each coro (see Coro::State).
459 - incompatible change: for very deep reasons, cede and cede_notself
460 cannot return anything, so nothing will be returned.
461 - possibly bring back 5.10 compatibility (untested).
462 - work around stupid (and wrong) warning on 5.10 :(.
463 - overlay the saved state over the context stack. This saves
464 a few hundred bytes per coroutine on average and also
465 speeds up context switching a bit.
466 - further tune default stack sizes.
467 - (more) correctly calculate stack usage in coro_rss.
468 - Coro::Storable::blocking_* did not properly lock
469 resulting in races between coroutines.
470 - added Coro::Storable::guard.
471 - stopping to trace a coroutine could destroy the cctx of
472 an unrelated coroutine.
473 - explain the relationship between Perl and C coroutines in
474 more detail in Coro::State.
475 - Coro::Util::inet_aton did not short-circuit dotted quad forms,
476 causing a fork per resolve. This also affected Coro::Socket.
477 - switch to a separate stack in $coro->call/eval to avoid
478 invalidating pointers.
479
480 4.03 Sat Oct 6 21:24:00 CEST 2007
481 - added Coro::throw method.
482 - minor code cleanups.
483
484 4.02 Sat Oct 6 02:36:47 CEST 2007
485 - fix a very minor per-coroutine memleak (a single codereference).
486 - fixed a bug where the currently in-use c context would be freed
487 prematurely (can happen only when programs change the stacksize
488 or use tracing).
489 - tracing can no longer keep a coro alive after it terminated.
490 - do static branch prediction in the common path for gcc. gives
491 about 2-5% speed improvement here.
492
493 4.01 Fri Oct 5 22:10:49 CEST 2007
494 - instead of recreating *a* standard output handle we simply
495 use STDOUT, which is faster and hopefully more robust.
496
497 4.0 Fri Oct 5 12:56:00 CEST 2007
498 - incompatibly changed Coro::Storable::freeze.
499 - major new feature: added Coro::Debug, for interactive coroutine
500 debugging, tracing and much more.
501 - major bug fix: unbelievable, but true: $_, $/ and many other
502 "saved" variables actually weren't being saved. This has been fixed,
503 of course, while increasing performance while losing all the save
504 flags.
505 - save flags are gone, and all the api functions dealing with them.
506 - added Coro::Semaphore::adjust.
507 - added Coro::Util::fork_eval.
508 - added Coro::Storable::{nfreeze,blocking_{freeze,nfreeze}}.
509 - added Coro::killall.
510 - reduce initial stack sizes to allow for "micro-coroutines".
511 - better async_pool resource management, moved parts of async_pool
512 handling to XS (major speed improvement).
513 - actually croak before modifying important data structures.
514 - refuse to transfer while compiling.
515 - possibly support eval EXPR better now.
516 - enable assembly per default on linux+bsd x86+amd64.
517 - all internal members were renamed _something for easier subclassing.
518 - many minor tweaks.
519
520 3.63 Wed May 16 14:10:06 CEST 2007
521 - implement handcoded assembly for x86/amd64 SVR ABI.
522
523 3.62 Fri Apr 27 21:36:06 CEST 2007
524 - upgrade libcoro (which might set unwind info correctly).
525 - change default on linux to setjmp/longjmp.
526
527 3.61 Thu Apr 19 12:36:18 CEST 2007
528 - Coro::Storable caused an endless loop when thawing invalid
529 pst's sometimes.
530 - use a Semaphore in Coro::Storable, as Storable doesn't
531 seem to be reentrant (although it is documented to
532 be threadsafe...).
533 - fix Coro::Signal to bring back the original unreliable
534 but stateful semantics.
535 - fixed a lot of typos in Coro.pm (patch submitted by David
536 Steinbrunner, which applied flawlessly).
537
538 3.6 Sat Apr 14 17:13:31 CEST 2007
539 - added some bugfixes to get eg/myhttpd working again.
540 - added Coro::Storable for often-cede'ing freeze/thaw.
541 - try to do a clean exit when a coroutine calls exit
542 (EXPERIMENTAL).
543 - got rid of indirect call through _coro_init.
544 - updated the partly antique examples in eg/ to
545 work again and be a bit less magic, too.
546 - fixed Coro::Signal semantics to work as documented again.
547
548 3.55 Sun Mar 25 01:20:47 CET 2007
549 - add SAVE_DEFFH to save the default filehandle and enable
550 it by default.
551 - finally move socket-operations from Coro::Socket to Coro::Handle
552 to be able to unblock foreign sockets.
553 - add Coro::State::save_also and guarded_save.
554 - add count accessor to Coro::Semaphore.
555 - add Coro::State::cctx_stacksize.
556 - just for the fun of it, do not rely on implicit context,
557 which can dramatically improve performance, but people
558 using windows-process-emulation perls do not care much about
559 performance.
560
561 3.51 Sun Mar 4 14:18:04 CET 2007
562 - fixed a problem when you weakened references to Coro::State's
563 (patch by Yuval Kogman).
564
565 3.501 Wed Feb 28 12:44:07 CET 2007
566 - rename some global symbols as macosx from hell redefines
567 them without asking.
568
569 3.5 Tue Feb 13 20:22:53 CET 2007
570 - do AnyEvent model detection earlier, avoiding problems
571 caused by first using AnyEvent and later Coro::Event.
572 - implement and document Coro::Event event objects.
573 - fix a potential problem in Coro::Event causing crashes.
574 - initialise PL_comppad when creating a new coroutine,
575 avoids crashes on early coro destruction.
576
577 3.41 Mon Jan 22 19:19:49 CET 2007
578 - readline on Coro::Handle did not support undefined $/,
579 nor did it deliver partial lines on EOF or error.
580 - implement malloc fallback for stack allocation because
581 stupid broken idiotic OSX has a stupid broken
582 idiotic fits-the-whole-os mmap "implementation" and
583 my dick feels longer if Coro is portable even to
584 obsolete platforms.
585
586 3.4 Fri Jan 19 21:52:54 CET 2007
587 - remove t/09_timer.t, as it isn't really testing much
588 but was rather flaky in practise.
589 - async_pool coro would keep arguments and callback alive until
590 it was reused.
591 - cancellation of a coroutine could cause spurious idle calls
592 in cede_notself.
593
594 3.3 Sat Jan 6 03:45:00 CET 2007
595 - implement $coro->on_destroy.
596 - Coro::Event blocking semantics have been changed,
597 documented and - hopefully - improved.
598 - fix nice adding, not subtracting, from priority.
599 - fix ->prio and api_is_ready (patch by Mark Hinds).
600 - fixed an assert ("... == prev__cctx->idle_te")
601 that could errronously trigger.
602 - fix various large and small memleaks.
603 - use a (hopefully) more stable cancel implementation
604 that immediately frees the coroutine data.
605 - cede/cede_notself return a status now.
606 - added Coro::guard function.
607 - added a global coroutine pool for jobs (on my machine,
608 I can create and execute 48k simple coros/s with async,
609 and 128k coros with async_pool).
610 - Coro::AIO now uses the coroutine priority as io priority.
611
612 3.2 Fri Dec 22 05:07:09 CET 2006
613 - improve portability to slightly older perls.
614 - use cleaner coroutine destruction.
615 - simplify configuration for users.
616 - optionally (unrecommended) prefer perl functions over
617 their coro replacements.
618
619 3.11 Tue Dec 5 13:11:24 CET 2006
620 - fixed some bogus assert's, but as perl.h disables assert even
621 without NDEBUG (thank you very much), not too many people should
622 notice (that did include myself). Andreas Kรถnig noticed, though :)
623 - do not save/restore PL_sortcxix on >= 5.9.x, it doesn't seem to have
624 it. Also noticed by Andreas Kรถnig :)
625 - save/restore tainted status.
626 - verified to pass the testsuite on my 5.9.5.
627
628 3.1 Mon Dec 4 23:03:40 CET 2006
629 - INCOMPATIBLE CHANGE: $/ is now per-coroutine (but slow).
630 - incompatible change: transfer flags are now per-state.
631 - give a better error message on deadlock.
632 - document Coro::nready.
633 - enhanced testsuite.
634
635 3.01 Sun Dec 3 23:47:42 CET 2006
636 - forgot to include Coro::Timer.
637
638 3.0 Sun Dec 3 22:57:25 CET 2006
639 - the "FINALLY COMPLETELY STABLE" release (coming soon:
640 the "FAMOUS LAST WORDS" release).
641 - implement a new stack sharing algorithm, which uses a stack
642 pool (size currently hardcoded).
643 - make stack sharing mandatory (it no longer uses a heuristic).
644 - eval/die no longer cause weird problems under heavy use.
645 - Coro::Event could cause livelocks if it was used but
646 no Coro::Event watchers were used.
647 - Coro::Event now uses asynccheck as prepare does not
648 check for changed watchers.
649 - Coro::Event allows multiple waiting coros and will wake up one
650 per event.
651 - Coro::Event should be cleaner and more efficient now.
652 - new utility function Coro::unblock_sub.
653 - document the sad fact that Event is no longer reentrant.
654 - putting a coroutine into the ready queue twice could under
655 some circumstances lead to stack corruption.
656 - minor incompatibility: subclassing Coro::State is supported
657 directly now without going through a _coro_state member.
658 - state/coro switching is much faster now.
659 - very minor optimisations and code/documentation cleanup.
660 - avoid problems due to compiler inlining.
661 - removed timers from Coro::Timer -> use AnyEvent instead.
662 - replaced Coro::idle coroutine by (cleaner) idle handler.
663 - updated to newest libcoro.
664 - implement enhanced support for valgrind.
665 - implement is_ready and return value for ready.
666 - removed Coro::Cont, it was a misnomer (it's generators),
667 and it was somewhat annoying to get it right. Will come back
668 if somebody asks for it :)
669 - many ->wait methods and Coro::Event could return
670 spuriously without the event having happened.
671
672 2.5 Tue Nov 7 12:22:33 CET 2006
673 - made Coro::Util, Coro::Select, Coro::Handle and Coro::Socket
674 use AnyEvent, moved them to Coro/.
675 - added Coro::LWP which contains all the uglyness required to
676 make LWP non-blocking.
677 - should work with perl 5.9.x now (Andreas Kรถnig made me do it).
678 - fixed another bug in Coro::Select when the timeout was undef.
679 - reuse PL_start_env for all coros, saves some memory per coroutine.
680 - manage PL_top_env differently, hopefully to avoid panic: top_env.
681 - timeout argument was not properly used in Coro::Socket.
682 - allow limited forms of subclassing in Coro::Handle/Coro::Socket.
683 - emulate undocumented(!) functionality of IO::Socket required
684 by LWP(!!).
685 - updated eg/lwp to work with newer lwp's.
686 - remove "FATAL: uncaught exception" prefix. Coroutines that die
687 kill the whole process, just as exceptions in the main "coroutine"
688 did already.
689
690 2.1 Wed Nov 1 23:01:13 CET 2006
691 - fix a long-standing bug in Coro::Select where select with
692 zero timeout would instead change the current default filehandle.
693 - use a simpler and hopefully more robust way to clone padlists
694 (uses less memory and a perl function instead of our own).
695 - coro can now create a stack guard area on many architectures.
696 - Coro::AIO properly reexports additional functions from IO::AIO.
697 - updated libcoro with a workaround for OS X,
698 pach and testing by Michael Schwern.
699
700 2.0 Tue Oct 24 05:47:17 CEST 2006
701 - support additional aio requests in Coro::AIO.
702
703 1.9 (never properly released due to a glitch)
704
705 1.8 Thu Feb 2 00:59:06 CET 2006
706 - applied suggested patch by SAMV to avoid problems during stupid
707 mark & sweep gc run.
708 - applied patch by Scott Walters for 5.9.3 compatibility.
709
710 1.7 Tue Dec 27 01:41:58 CET 2005
711 - added Coro::AIO, a thin wrapper around IO::AIO.
712 - improved Makefile.PL explanations.
713
714 1.6 Mon Dec 12 21:30:05 CET 2005
715 - additionally save PM_curpm might fix as-of-yet
716 unseen problems with regex matches being attributed to the
717 wrong package.
718 - add t/10_bugs.t, which currently checks against imho broken
719 perls that use 0.26MB of stack space per Perl_magic_get
720 invocation (newer linuxes) as opposed to the 0.0002MB perl
721 normally uses.
722 - make stacksize configurable for the ultimate debian experience.
723
724 1.51 Mon Dec 12 18:48:36 CET 2005
725 - remove debugging warn() accidentally left in Coro::Select.
726
727 1.5 Tue Nov 29 13:32:44 CET 2005
728 - use Coro::Event inside Coro::Select to avoid spurious deadlocks.
729 - fix Coro::Select 'select'.
730 - strict'ify some modules.
731 - libcoro errornously restored the SIGUSR2 handler to SIGUSR1.
732 - use XSLoader in selected modules.
733 - remove some 5.6 compatibility cruft.
734
735 1.4 Tue Sep 6 00:11:05 CEST 2005
736 - libcoro did not take into account the trampoline on amd64,
737 when 'l'inux method was used.
738
739 1.31 Tue Aug 30 23:31:33 CEST 2005
740 - some portability fixes/workarounds.
741
742 1.3 Sat Aug 20 03:08:56 CEST 2005
743 - no code changes. module seems to work fine.
744 - improve Coro::State docs, remove reference to nonexisting
745 Coro::State::flush method.
746 - no longer autodetect windows, present it as an option instead.
747
748 1.2 Mon May 16 02:00:55 CEST 2005
749 - included libcoro.c earlier in State.xs, might improve portability.
750 - use a faster and possibly more stable Coro::Cont implementation.
751 - accept x86_64 in addition to amd64 for optimized linux-amd64 support.
752 - fix bugs Coro::SemaphoreSet that could cause locks never to be freed.
753 - fix bugs in CoroAPI.h and document it in Coro::MakeMaker.
754
755 1.11 Thu Mar 3 18:00:52 CET 2005
756 - change of contact address.
757
758 1.1 Tue Feb 22 20:51:16 CET 2005
759 - support [l]inux method on amd64.
760 - allow some context switches while compiling/eval'ing.
761
762 1.0 Tue Aug 10 03:47:27 CEST 2004
763 - Event 0.89 fixes the bug again.
764 - fix the remaining(??) memleak. I didn't free the cache-AV for
765 padlists, so one AV per created anon-sub leaked.
766
767 0.97 Fri May 14 15:23:32 CEST 2004
768 - remove stupid gcc extension that was an experiment only anyway.
769
770 0.96 Thu May 13 18:09:29 CEST 2004
771 - tell people if they have yet another broken Event variant.
772 - made Coro to work _better_ with threads (it seems that, as long
773 as you use Coros in one thread only everything will work fine.
774 Looking at the perl source the problem might be missing locking
775 between Coro and some perl internal routines. That's not easily
776 fixable).
777 - (hopefully) got rid of the static xnext inside transfer.
778
779 0.95 Sun Feb 15 03:20:28 CET 2004
780 - removed Coro::State::flush, implemented a new and much
781 faster caching mechanism by attaching magic to the CV.
782 - a memleak with (real) closures remains.
783 - dramatically reduced size of stacks to 65536 (linux/x86), after
784 observing that even myhttpd never had more than a kilobyte
785 of stack in use (stack is still only physically allocated in
786 pagesize increments for systems suporting mmap).
787 - fix for a "cygwin-64" architecture added, proposed by Stefan Traby.
788 - fix for threaded perl.
789
790 0.9 Sun Nov 23 04:32:39 CET 2003
791 - fix a few very hard to track down but (of course) stupid
792 memory leaks.
793 - stringify version number to avoid locale problems :(.
794 - perl 5.9.0 does not have SvPADBUSY, reported by Scott Walters,
795 hopefully ignoring it "just works".
796
797 0.8 Wed Nov 5 19:38:40 CET 2003
798 - port to perl5.005_03, but only 5.8.x+ is supported!
799 - honor LocalAddr even without LocalPort in Coro::Socket.
800
801 0.7 Tue May 27 03:12:38 CEST 2003
802 - the version jump indicates some level of testing, not gobs
803 of new features.
804 - uh... I found the prompt function in ExtUtils::MakeMaker.
805 Highly correct stuff, that is...
806 - fixed(?) a bug with die's in coroutines causing "panic: top_env".
807 the fix is not well understood by the author, so beware :(.
808
809 0.652 Thu May 8 02:54:46 CEST 2003
810 - Applied patch by Slaven Rezic to set default to "s" on FreeBSD,
811 cause version 4 doesn't have ucontext.
812 - Benjamin Reed reported that setjmp works fine on darwin,
813 so preselect it.
814
815 0.651 Sat Mar 29 15:00:23 CET 2003
816 - fix a bug in Coro::Handle where some bytes could get lost
817 on reading, reported by jason@nichego.net.
818
819 0.65 Sun Mar 23 00:08:26 CET 2003
820 - added a README.linux-glibc.
821 - new module Coro::Select.
822 - also save/restore PL_comppad, fixes segfaults with 5.8.1.
823 (I never claimed that I know what I am doing ;).
824 - set default to setjmp/longjmp on non-x86-linux.
825
826 0.6 Thu Nov 21 11:09:06 CET 2002
827 - made lazy context switching the default.
828 - fixed the bug where SemaphoreSet::guard
829 was the same as timed_guard and timed_guard was missing.
830 - fix a memleak in Coro::Timer.
831 - reclassified context sharing as not experimental.
832
833 0.534 Sun Apr 14 03:05:12 CEST 2002
834 - fixed backspace => tab in header parsing (myhttpd).
835 - added eg/readline, for elmex.
836 - Coro::Event now calls ->start on first call instead of
837 ->again, which created wrong timeouts for the first ->next.
838 - fixed a bug where Coro::Socket returned a closed socket instead
839 of nothing on ECONNREFUSED and similar error conditions.
840
841 0.533 Mon Feb 18 18:49:40 CET 2002
842 - adapt to finally working Event-0.85.
843 - fixed buggy event initialization (trapped under windows, but how
844 could it possibly have worked anywhere?)
845
846 0.532 Wed Jan 16 02:45:32 CET 2002
847 - removed the reference to pp_entersub, might work on
848 (some) win32 perls now (testsuite works under cygwin,
849 without Event, which freezes).
850
851 0.531 Mon Dec 10 22:18:44 CET 2001
852 - Fixed a bug in SemaphoreSet::timed_down.
853
854 0.53 Tue Nov 27 21:11:13 CET 2001
855 - some tests for Event.
856 - slightly different internal architecture to get refcnt into the object
857 and not the reference. might fix a few bugs, certainly did introduce
858 new ones.
859 - Coro::Timer (independent of Event).
860 - new timed_wait functions for Coro::Signal, Semaphore, SemaphoreSet.
861
862 0.52 Tue Nov 6 21:36:18 CET 2001
863 - ported to cygwin (trivial).
864 patch by Gerrit P. Haase <gp@familiehaase.de>.
865 - small setjmp code fix by Sullivan.DanielJ@epamail.epa.gov.
866
867 0.51 Thu Nov 1 20:39:01 CET 2001
868 - terminate/cancel now work properly (otherwise termination
869 could cause a "next coroutine is not and contains not..."
870 error.
871 - added Coro::Socket::shutdown.
872 - Coro::Event::loop no is the same as Event::loop.
873 - implemented terminate with args + join.
874
875 0.5 Fri Sep 28 16:15:35 CEST 2001
876 - fixed "print" on a Coro::Handle. The print method worked.
877 - small tweaks (seem to reduce memory consumption a lot)
878 in various modules.
879 - splendid use of "no warnings" scattered throughout.
880 - added Coro::Handle::rbuf, fixed Coro::Handle::fh.
881
882 0.49 Sun Sep 16 02:42:45 CEST 2001
883 - changed some method calls to function calls for speed
884 inside Coro::Handle.
885 - make Coro::Handle use an array instead of a hash for
886 speed reasons.
887 - IRIX mystery solved: it's SGI's NT, after all: "standard, huh?".
888 sjlj and ucontext should now work.
889 - IRIX-specific port for libcoro.
890 - swapped order of accept results to match IO::Socket.
891 - changed getsock/peername to sock/peername to match IO::Socket.
892 - fixed a bug that caused segfault when returning to main task
893 under some circumstances.
894 - other bugfixes.
895
896 0.45 Sun Sep 2 02:54:01 CEST 2001
897 - new method Coro::Handle::timeout.
898 - corrected speling of set...name to get...name in Coro::Socket.
899 - Coro::Socket::accept now returns a Coro::Socket, not a Coro::Handle.
900 - Coro::Handle now supports fileno and FILENO.
901 - added eg/myhttpd, a web server.
902 - Coro::Socket now accepts numeric ports as well as "port(number)"
903 syntax.
904 - moved some scheduling primitives into xs code. more to come.
905 - new simple priority system.
906 - implemented Coro::Event using XS, almost four times faster!
907 - small memory corruption problem fixed (boy that was difficult).
908
909 0.13 Wed Aug 8 16:53:07 CEST 2001
910 - new method Coro::Handle::readline.
911 - added eg/lwp and Coro::Handle::autoflush, to make LWP work.
912 (see eg/lwp on how to make LWP non-blocking).
913 - renamed Coro::Socket::new_inet to new.
914 - added Coro::Util, some utility functions.
915 - die/eval now works better (still get annoying (but true ;)
916 "Callback called exit" messages).
917 - Coro::Handle now supports timeout, and Coro::Socket Timeout.
918 - much better Coro::Socket support.
919
920 0.12 Fri Jul 27 04:19:01 CEST 2001
921 - do not use mmap if neither MAP_ANON nor MAP_ANONYMOUS is defined.
922 - very experimental stack sharing algorithm. not 100% safe but
923 should work well in practise ;)
924 - added Coro::RWLock.
925 - Coro::Specific now works.
926
927 0.11 Tue Jul 24 22:49:21 CEST 2001
928 - added specialized hack for newer and older linux versions (fast).
929 - renamed Coro::Event::IO to Coro::Handle.
930 - new module Coro::Socket.
931
932 0.1 Tue Jul 24 01:47:53 CEST 2001
933 - release candidate 3. A rather new internal structure :(
934 - the great renaming: Damian Conway gave me a suitable replacement
935 for yield.
936 - added Coro/libcoro, a portable coroutine implementation for C
937 (not even perl-dependent!!!), which can be used by Coro::State.
938 - renamed SAVE_DEFSV (on xs-level only) to avoid symbol clash in
939 perl-5.6.
940 - new function Coro::Event::idle.
941 - the idle process is now overriden
942 by default in Coro::Event.
943 - Coro::Channel now enforces the size.
944 - canceling events now works in all cases.
945 - Coro state now include $_ and $@.
946 - yet another bug workaround that I do not really understand :(
947 - new module Coro::Event::IO (very undocumented) to do
948 non-blocking i/o.
949 - performance tuning.
950
951 0.08 Thu Jul 19 06:13:25 CEST 2001
952 - release candidate 2 (new functionality).
953 - add Coro::State::flush function.
954 - transfer now supports a flags value (mostly for speed ATM).
955 - might compile and work in the presence of threads now.
956 - continuations have a saner syntax.
957 - no more memleaks.
958
959 0.07 Tue Jul 17 17:40:18 CEST 2001
960 - release candidate 1 ;)
961 - slightly nicer code.
962 - fixed a scheduling bug in Coro::Event.
963
964 0.06 Tue Jul 17 04:23:24 CEST 2001
965 - ok, I found the showstopper - the same sub must not be
966 re-entered in two different coroutines, otherwise => crash. I
967 see no easy solution to this problem, except by walking the call
968 chaing and saving/restoring the cv's, which is what I do now.
969 - memory leaks still latent, especially at thread termination.
970 - Coro::Event now works (haha).
971
972 0.05 Sun Jul 15 17:32:20 CEST 2001
973 - fixed a few issues in Coro::Event.
974 - I forgot to include Coro::Event in 0.04 :(:(:(
975
976 0.04 Sun Jul 15 05:24:59 CEST 2001
977 - @_ is now properly localized.
978 - Coro::State is now easier subclassable.
979 - Coro::Cont now coroutine-aware.
980 - Coro::Specific is a low-overhead module to create
981 coroutine-specific vars.
982 - Coro::Event provides a simple interface to Event.
983
984 0.03 Fri Jul 13 14:51:52 CEST 2001
985 - transfer() now implemented in XS (beware).
986 - new module Coro::Cont for really faked continuations.
987 - big internal architecture changes: Coro::State is now
988 really low-level and can thus be used to implement other
989 interesting things, While "Coro::" implements a process-like
990 model. Still crude and subject to change.
991 - $_ and $@ are no longer being localized.
992
993 0.02 Tue Jul 10 01:38:17 CEST 2001
994 - implemented "async" attribute.
995 - $_ and $@ are now localized.
996 - added Coro::Channel.
997 - more testcases, still no docs.
998
999 0.01 Tue Jul 3 02:18:41 CEST 2001
1000 - original version; copied from Convert::Scalar.
1001